One national programme offered two gastric cancer tests. Only one of them changed who died

Korea launched national gastric cancer screening in 1999, offering everyone aged 40 and over a choice between two tests every two years: upper endoscopy, or an upper gastrointestinal series done with barium and X-rays. Both were in the programme. Both were paid for. Eighteen years of national data show that only one of them was doing anything.

The comparison the programme accidentally ran#

A study published in Gastroenterology used the programme's own records as a natural experiment. From a cancer-free cohort of 16,584,283 people, it identified 54,418 individuals newly diagnosed with gastric cancer between 2004 and 2009 who died before December 2012, and matched each to four controls alive on the date the case died, matched on year of entry, age, sex and socioeconomic status.

An odds ratio of 0.98 with a confidence interval from 0.95 to 1.01 is not a weak effect. It is the absence of one, measured precisely, in a national programme that offered the test for eighteen years.

The pattern that argues this is causal#

The obvious objection to any observational screening study is that people who attend screening are different from people who do not — healthier, wealthier, more health-conscious — and that the difference, not the test, produces the result. The strongest evidence against that reading here is not a statistical adjustment. It is the shape of the endoscopy result.

• One endoscopic screen: odds ratio 0.60 (95% CI 0.57 to 0.63).
• Two: 0.32 (0.28 to 0.37).
• Three or more: 0.19 (0.14 to 0.26).

Health-conscious attendance would be expected to raise both figures together. Instead the benefit deepens with each additional endoscopy while the barium series stays at unity no matter how often it is repeated. A confounder that produces a dose-response in one arm of the same programme and none in the other would have to be a very unusual confounder.

What a clear scan is worth#

A second national cohort, of 15,850,288 people followed to 2017 for cancer and 2019 for death, asked a different question: what does a negative screening result predict? Over that period it recorded 230,783 upper gastrointestinal cancers and 99,348 deaths from them.

Both negative results carry reassurance. A clear endoscopy carries much more of it. The largest reductions in both studies fell in the 60-to-69 age band, where a negative endoscopy was followed by a hazard ratio of 0.19 for upper gastrointestinal cancer death.

What this settles and what it does not#

These are observational analyses of a national programme, not randomised trials, and neither can fully exclude that the people choosing endoscopy differed from those choosing barium in ways the matching did not capture. What makes the comparison unusually strong is that both groups were inside the same programme, invited the same way, at the same intervals, for the same disease.

Neither study measures the harms of endoscopy — perforation, bleeding, sedation events, or the consequences of findings that would never have caused illness. A test that halves gastric cancer mortality is not thereby free, and the studies that would quantify what it costs are not these.

For somebody reading a checkup package, the practical content is narrow and firm. Where a Korean programme offers a choice between endoscopy and a barium upper gastrointestinal series for gastric cancer, the national data say those are not two versions of the same test.